Multi-constraint fusion agv path planning method and related components

Abstract

This invention provides a multi-constraint fusion AGV path planning method and related components, including: constructing a topological road network containing production process constraints to obtain a topological road network containing mandatory nodes; performing constraint filtering on the topological road network based on the AGV's physical parameters to obtain a feasible topological road network; using a search algorithm to perform process-oriented global path search in the feasible topological road network to generate an initial globally optimal path; collecting environmental data in real time and performing dynamic obstacle detection based on the environmental data to obtain obstacle information; performing local replanning on the initial globally optimal path based on the obstacle information to generate a final executable path; smoothing the final executable path and converting it into control commands, and driving the AGV to execute through the control commands. Compared with the prior art, this invention improves the executability of path planning.

Description

Technical Field

[0001] This invention relates to the field of automated guided vehicle (AGV) technology, specifically to an AGV path planning method and related components that integrate multiple constraints. Background Technology

[0002] Automated Guided Vehicles (AGVs), as intelligent material handling equipment, are widely used in industrial production scenarios such as intelligent warehousing and flexible manufacturing. Path planning is the core technology to ensure the efficient and safe operation of AGVs. Existing AGV path planning methods typically employ search algorithms such as A* and Dijkstra's algorithm, using the shortest path distance or time as a single optimization objective to plan an optimal path from the starting point to the destination in a static environment. However, existing AGV path planning technologies only consider single constraints such as distance or time, leading to poor executability of the planned paths. Summary of the Invention

[0003] The purpose of this invention is to provide an AGV path planning method and related components that integrate multiple constraints, thereby improving the executability of path planning.

[0004] In a first aspect, the present invention provides an AGV path planning method with multi-constraint fusion, comprising: constructing a topological road network containing production process constraints to obtain a topological road network containing mandatory nodes; performing constraint filtering on the topological road network based on the physical parameters of the AGV to obtain a feasible topological road network; performing a process-oriented global path search in the feasible topological road network using a search algorithm to generate an initial globally optimal path; collecting environmental data in real time, performing dynamic obstacle detection based on the environmental data to obtain obstacle information; performing local replanning on the initial globally optimal path based on the obstacle information to generate a final executable path; smoothing the final executable path and converting it into control instructions, and driving the AGV to execute through the control instructions.

[0005] According to one embodiment of the present invention, constructing a topological road network containing production process constraints to obtain a topological road network containing mandatory nodes includes: acquiring workshop environment data, generating a grid map based on the workshop environment data, and extracting key nodes from the grid map to construct an initial topological node set; establishing topological edges connecting adjacent nodes in the initial topological node set, and setting distance attributes, turning angle attributes, and process weight attributes for each topological edge; parsing production process data, extracting a set of necessary process points, and generating a process chain according to the process sequence; mapping each point in the set of necessary process points to the grid map, and determining them as mandatory nodes; merging the mandatory nodes, the initial topological node set, and the topological edges to obtain the topological road network.

[0006] According to one embodiment of the present invention, the constraint filtering of the topological road network based on the physical parameters of the AGV to obtain a feasible topological road network includes: obtaining the physical parameters of the AGV, the physical parameters including a minimum turning radius and a maximum travel speed; traversing all topological edges in the topological road network and calculating the turning angle corresponding to each topological edge; comparing the turning angle of each topological edge with the minimum feasible turning angle calculated based on the minimum turning radius, and discarding topological edges with turning angles smaller than the minimum feasible turning angle; evaluating the travel time of the remaining topological edges based on the maximum travel speed to obtain an estimated speed, and discarding topological edges with estimated speeds exceeding the maximum travel speed; and performing redundancy removal processing on the nodes and edges in the filtered topological road network, deleting duplicate nodes and redundant edges to obtain the feasible topological road network.

[0007] According to one embodiment of the present invention, the step of using a search algorithm to perform process-oriented global path search in the feasible topology network to generate an initial globally optimal path includes: setting multi-dimensional weight coefficients for the search algorithm, wherein the weight coefficients include a first coefficient for calculating turning costs, a second coefficient for calculating process costs, and a third coefficient for calculating safety costs; decomposing the global task into multiple sub-path search tasks based on the process chain, with adjacent process necessary points as starting and ending points; for each sub-path search task, starting from the starting point, calculating the total cost of each candidate node based on a multi-constraint weighted heuristic function; guiding the search according to the total cost until reaching the end point of the sub-path to obtain the optimal sub-path; and concatenating all the optimal sub-paths in the process chain order to generate the initial globally optimal path.

[0008] According to one embodiment of the present invention, the real-time acquisition of environmental data for dynamic obstacle detection to obtain obstacle information includes: acquiring raw environmental perception data collected by sensors mounted on the AGV; performing data fusion processing on the raw environmental perception data to identify dynamic obstacles in the environment; extracting the position coordinates, movement speed, and movement direction of the dynamic obstacles; identifying whether the dynamic obstacles have entered the preset perception range of the AGV and affected the current travel path; when the identification result is yes, acquiring the obstacle information and generating a warning signal; wherein, the obstacle information includes the position coordinates, movement speed, and movement direction of the dynamic obstacles.

[0009] According to one embodiment of the present invention, the step of performing local replanning on the initial global optimal path based on the obstacle information to generate a final executable path includes: locating the affected topological edges in the feasible topological network according to the obstacle information; marking the affected topological edges as temporarily impassable and adjusting their process cost and safety cost; determining a local replanning region using a preset number of nodes before and after the affected road segment as the start and end points of the local search; performing path search within the local replanning region using a local replanning heuristic function; replacing the affected road segments in the initial global optimal path with the searched local optimal path, and concatenating them to generate the final executable path.

[0010] Secondly, the present invention also provides a multi-constraint fusion AGV path planning device, comprising: a topology network construction module for constructing a topology network including production process constraints to obtain a topology network including mandatory nodes; a constraint filtering module for performing constraint filtering on the topology network based on the physical parameters of the AGV to obtain a feasible topology network; a path search module for performing process-oriented global path search in the feasible topology network using a search algorithm to generate an initial globally optimal path; a dynamic obstacle detection module for collecting environmental data in real time and performing dynamic obstacle detection based on the environmental data to obtain obstacle information; a dynamic replanning module for performing local replanning on the initial globally optimal path based on the obstacle information to generate a final executable path; and a path execution module for smoothing the final executable path and converting it into control commands, and driving the AGV to execute through the control commands.

[0011] Thirdly, the present invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, it implements the multi-constraint fusion AGV path planning method of the above embodiments.

[0012] Fourthly, the present invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the multi-constraint fusion AGV path planning method of the above embodiments.

[0013] The present invention produces at least the following beneficial effects: This invention achieves a deep integration of path planning with production process flow and AGV physical characteristics by constructing a topological road network that includes production process constraints and integrating heuristic functions with multi-dimensional constraints, thereby improving the executability of path planning. Attached Figure Description

[0014] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0015] Figure 1 This is a flowchart of the multi-constraint fusion AGV path planning method according to an embodiment of the present invention; Figure 2 yes Figure 1 A schematic diagram of a specific implementation method for step S10; Figure 3 yes Figure 1 A schematic diagram of a specific implementation method for step S20; Figure 4 yes Figure 1 A schematic diagram of a specific implementation method for step S30; Figure 5 yes Figure 1 A schematic diagram of a specific implementation of step S40; Figure 6 yes Figure 1 A schematic diagram of a specific implementation method for step S50; Figure 7 yes Figure 1 A schematic diagram of a specific implementation method for step S60; Figure 8 This is a schematic block diagram of an AGV path planning device with multi-constraint fusion according to an embodiment of the present invention. Detailed Implementation

[0016] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0017] It should be understood that, when used in this specification and the appended claims, the terms “comprising” and “including” indicate the presence of the described features, integrals, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more of its features, integrals, steps, operations, elements, components and / or collections thereof.

[0018] It should also be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0019] It should also be further understood that the term "and / or" as used in this specification and the appended claims refers to any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0020] To facilitate the description of this application, the following concepts related to this application are introduced.

[0021] Automated Guided Vehicle (AGV): A transport vehicle equipped with an automatic guidance system, capable of traveling along a prescribed guidance path, and possessing safety protection and various transfer functions. In this application, the AGV is the main entity for path planning and execution, and its physical parameters (such as minimum turning radius and maximum speed) are key constraints on path feasibility.

[0022] Topological network: A graph structure consisting of nodes and edges, used to abstract and represent path networks in a physical environment. In this application, the topological network consists of key workshop nodes (such as workstations and turning points) and topological edges connecting them, and its attributes (such as distance, turning angle, and process weight) carry multi-dimensional information required for path planning.

[0023] Heuristic function: In search algorithms (such as the A* algorithm), a function is used to evaluate the estimated cost from the current node to the target node. In this application, the heuristic function is improved to a weighted sum of multiple constraints such as distance, turning, process weights, and safety, which is used to guide the search process to generate a better and executable path.

[0024] B-spline curves: a mathematical method for curve modeling and interpolation, characterized by local support and convex hull properties. In this application, cubic B-spline curves are used to smooth discrete planning paths, generating continuous trajectories with gentle curvature changes to adapt to the kinematic characteristics of AGVs.

[0025] Process flow data refers to structured data describing the sequence of each process step in product manufacturing, as well as the required materials, equipment, and workstations, such as process route diagrams and process flow cards. In this application, the process flow data is parsed to extract the set of essential process points, generating a process chain to guide global path search and ensure that path planning conforms to the production process priority.

[0026] Raster map: A map representation method that divides the environment into raster cells of equal size, with each raster cell storing environmental information (such as obstacles and free space).

[0027] Jump Point Search (JPS) is an optimization algorithm for path planning on grid maps. It significantly reduces the number of search nodes by identifying and skipping symmetrical nodes on the path and retaining only key turning points (jump points) for expansion.

[0028] This application provides an AGV path planning method with multi-constraint fusion, referring to... Figure 1 ,include: S10: Construct a topological road network containing production process constraints, resulting in a topological road network with mandatory nodes. The purpose of this step is to digitize the physical workshop environment and production process flow, constructing a road network model that reflects spatial connectivity and embeds process priorities, providing a data foundation for subsequent path searching. Specifically, CAD drawings of the workshop layout and on-site scanning data can be collected using industrial-grade scanners or LiDAR to generate a grid map for accurately representing obstacles. Simultaneously, key nodes such as workstations, charging stations, and turning points are extracted from the grid map to construct an initial topological node set, and topological edges connecting adjacent nodes are defined, initializing their distance, turning angle, and other attributes. Further, process flow data is parsed, and essential process points such as processing and quality inspection are extracted as a set of essential process points. Process chains are generated according to the process sequence, and these points are mapped to the topological road network and marked as mandatory nodes, ensuring that path planning must pass through these nodes. S20: Based on the physical parameters of the AGV, constrain the topological network to obtain a feasible topological network. The purpose of this step is to eliminate path segments in the topological network that do not conform to the kinematic constraints of the AGV itself, ensuring that the planned path is physically executable. Specifically, core physical parameters such as the minimum turning radius R and maximum travel speed Vmax can be obtained from the AGV controller. Each edge in the topological network is traversed, its turning angle is calculated, and compared with the minimum feasible turning angle calculated based on the minimum turning radius. Edges with angles too small and impossible for the AGV to complete are eliminated. Simultaneously, the estimated travel speed of the AGV on each edge is evaluated, and edges exceeding the maximum speed limit are eliminated. Finally, algorithms such as jump point search are used to simplify the filtered network, removing duplicate or meaningless intermediate nodes to obtain a simplified and feasible topological network. S30: A search algorithm is used to perform a process-oriented global path search in the feasible topology network to generate an initial globally optimal path. The purpose of this step is to find a globally optimal path within the feasible network that satisfies physical constraints, one that follows the production process sequence while also being optimal overall. Specifically, weight coefficients of the improved search algorithm can be configured, such as turning cost weight α, process cost weight β, and safety cost weight γ. Based on the process chain, the global transportation task is decomposed into multiple sub-path search tasks, each starting and ending at a point that is a necessary step in an adjacent process. For each sub-task, starting from the starting point, the total cost f(n) of each candidate node is calculated using a heuristic function that incorporates multiple constraints. This cost integrates the actual travel distance, heuristic distance, turning cost, process weight cost, and safety cost. The algorithm guides the search based on the principle of minimizing the total cost until the sub-path endpoint is reached, obtaining the optimal sub-path. Finally, all optimal sub-paths are concatenated according to the process chain sequence to form the initial globally optimal path. S40: Real-time acquisition of environmental data, dynamic obstacle detection based on the environmental data, and obtaining obstacle information. The purpose of this step is to perceive the dynamic changes in the AGV's operating environment and provide a basis for real-time obstacle avoidance decisions. Specifically, raw perception data of the surrounding environment can be acquired in real time using the lidar and vision sensors mounted on the AGV. This data is processed using a sensor fusion algorithm to identify dynamic obstacles such as pedestrians, other vehicles, or temporary stacks of objects, and to extract their position coordinates, speed, and direction of movement. The system continuously determines whether these obstacles have entered the AGV's preset perception range (e.g., within 5 meters) and may affect the current planned path. If the determination result is yes, obstacle information containing detailed obstacle information is generated, and a warning signal is triggered. S50: Based on the obstacle information, perform local replanning on the initial global optimal path to generate a final executable path. The purpose of this step is to quickly and efficiently adjust the path when the global path is disturbed by dynamic obstacles, avoiding the excessive time consumption caused by global replanning. Specifically, based on the obstacle information, locate the topological edges covered or affected by obstacles in the feasible topological network, mark them as temporarily impassable, and significantly increase their process cost and safety cost weights. Use a preset number of nodes (e.g., 5) before and after the affected road segment as the new starting and ending points of the local search to determine a local replanning region. Within this region, use a local replanning heuristic function with an added conflict cost term to re-search for the path. Replace the affected road segments in the initial global path with the locally optimal path obtained from the search, and concatenate them to generate a new, safe, and ultimately executable path. S60: The final executable path is smoothed and converted into control commands, which then drive the AGV to execute. The purpose of this step is to transform the discrete topological path into a continuous, smooth sequence of commands that meets the motion control requirements of the AGV, ensuring that the AGV can smoothly and accurately track the path. Specifically, a series of key path points can be extracted from the final executable path as control points. A cubic B-spline curve algorithm is used to perform interpolation calculations based on these control points, generating a smooth, continuous curve and eliminating sharp turns in the path. The smoothed curve is then discretized to obtain a series of dense path point sequences. Combined with the AGV's kinematic model, the linear velocity and angular velocity commands that the AGV should execute at each sampling moment are calculated. Finally, these control commands are sent to the AGV's motion controller, driving the AGV to travel along the planned path and complete the task.

[0029] The embodiments of this application employ the above method, which produces at least the following beneficial effects: By constructing a topological road network that incorporates production process constraints and integrating a heuristic function with multi-dimensional constraints, a deep integration of path planning with production process flow and AGV physical characteristics is achieved. This method solves the problem of poor path executability caused by existing search algorithms that only consider a single constraint, thus generating paths that combine optimality and executability.

[0030] By implementing constraint filtering and path smoothing based on physical parameters, kinematic constraints on AGVs are achieved. This method solves the problem of physically infeasible planned paths, such as sharp-angle turns and speed exceeding limits, thereby improving the stability and safety of AGV operation.

[0031] By employing dynamic obstacle detection and a local replanning mechanism, a rapid response to environmental changes is achieved. This method solves the problem of excessively long global replanning time and its impact on real-time operation, thereby improving the operational efficiency and robustness of AGVs in dynamic environments.

[0032] This method can be applied to intelligent warehousing and logistics scenarios, providing AGVs with efficient material transfer path planning. Based on the warehouse layout and inbound / outbound processes, this method can automatically plan the optimal transfer path that meets the AGV's turning capabilities and avoids congested areas. It can also quickly adjust when encountering pedestrians or other vehicles, ensuring the continuity and safety of warehousing operations.

[0033] This method can be applied to flexible manufacturing workshops, providing process-oriented material delivery path planning for AGVs. It strictly follows the production process sequence, ensuring materials are delivered to each workstation on time and in order, while also considering personnel safety and equipment avoidance within the workshop, thus improving the overall collaborative efficiency of the flexible production line.

[0034] In some embodiments, reference Figure 2 Step S10 includes: S11: Acquire workshop environmental data, generate a raster map based on the workshop environmental data, and extract key nodes from the raster map to construct an initial topological node set. The purpose of this step is to discretize the continuous physical space into a computer-processable map model and extract key decision points for path planning. The specific implementation steps are as follows: The first step is data acquisition and map generation. For example, the workshop is scanned using LiDAR SLAM technology to generate a grid map with an accuracy of 0.1m, where each grid is marked as passable or an obstacle.

[0035] The second step is key node extraction. For example, image processing algorithms are used to identify workstation center points, passageway intersections, turning points, etc., in the raster map, and their coordinates are entered into the initial topology node set V = {v1, v2, ..., v n}

[0036] S12: Establish topological edges connecting adjacent nodes in the initial topological node set, and set distance, turning angle, and process weight attributes for each topological edge. The purpose of this step is to define the connectivity between nodes and assign them the attribute values ​​required for planning. The specific implementation steps are as follows: The first step is to construct the topological edge. For example, for node pairs in the node set V whose Euclidean distance is less than the threshold and whose paths are traversable, a topological edge E = {e_ij | v_i, v_j ∈ V} is established to connect adjacent nodes in the initial topological node set.

[0037] The second step is attribute initialization. For example, the actual length of the topological edge e_ij is calculated as the distance attribute; the angle between the topological edge e_ij and the adjacent edges is calculated as the turning angle attribute; according to the process flow data, if the topological edge e_ij is located on the core process path, its process weight attribute is initialized to 0.1, otherwise it is initialized to 0.9.

[0038] S13: Analyze the production process data, extract the set of necessary process points, and generate a process chain according to the process sequence. The purpose of this step is to transform the production flow into a mandatory constraint sequence for path planning. The specific implementation steps are as follows: The first step is data parsing, such as parsing process flow data (e.g., process route diagrams, process flow cards) to identify essential process nodes such as raw material area, processing area, quality inspection area, and storage area. The essential process node set P = {p1, p2, ..., p...} is then extracted. m}

[0039] The second step is process chain generation. The extracted set of necessary process points P is arranged according to the production flow sequence to form a process chain C = [p1→p2→...→p m].

[0040] S14: Map each point in the set of necessary process points to the grid map and identify them as mandatory nodes. Merge the mandatory nodes, the initial topology node set, and the topology edges to obtain the topology network. The purpose of this step is to embed process constraints into the topology network structure. The specific implementation steps are as follows: The first step is node mapping and labeling, for example, mapping the coordinates of nodes in process chain C to a grid map and labeling their attributes as forced nodes in the topology node set.

[0041] The second step is data merging, which involves merging the forced nodes with the initial topology node set, updating the node set V, and retaining all topology edges e and their attributes, ultimately obtaining the complete topology network structure.

[0042] By mapping essential process points to mandatory nodes and generating process chains, this invention achieves a mandatory binding between path planning and production process flow. This method solves the problem that traditional path planning may skip critical processes, leading to production interruptions, thereby ensuring that AGV operations follow the production rhythm and improving the continuity and controllability of the production process.

[0043] In some embodiments, reference Figure 3 Step S20 includes: S21: Obtain the physical parameters of the AGV, including the minimum turning radius and maximum travel speed. The purpose of this step is to obtain the kinematic limit parameters of the AGV itself, as a basis for judging the feasibility of the path. The specific implementation steps are as follows: The first step is to read the parameters, such as the AGV's minimum turning radius Rmin and maximum driving speed Vmax, through the AGV's CAN bus or configuration file.

[0044] The second step is parameter storage, for example, storing the read parameters in the configuration file of the path planning module for subsequent constraint calculations.

[0045] S22: Traverse all topological edges in the given road network and calculate the turning angle corresponding to each topological edge. The purpose of this step is to quantify the turning requirements of each path segment. The specific implementation steps are as follows: The first step is angle calculation. For example, for a topological edge e_ij, which connects nodes v_i and v_j, calculate the angle θ_ij between vector v_i-1→v_i and vector v_i→v_j.

[0046] The second step is angle storage, for example, updating the calculated θ_ij value to the attributes of the topological edge e_ij.

[0047] S23: Compare the turning angle of each topological edge with the minimum feasible turning angle calculated based on the minimum turning radius, and discard topological edges whose turning angle is less than the minimum feasible turning angle. The purpose of this step is to eliminate sharp turning paths that the AGV cannot complete. The specific implementation steps are as follows: The first step is to determine the minimum feasible turning angle based on the minimum turning radius. For example, obtain the minimum turning radius R_min of the AGV, and calculate the minimum radius of curvature required for the AGV to pass through the turn based on the path geometry (including the length and angle of the straight segments before and after the turn). If the required radius of curvature is greater than R_min, the turn is determined to be impassable. To avoid complex calculations, a simplified judgment can be adopted: when the turning angle θ_ij of the topological edge is less than a preset threshold (e.g., 30°), the turn is considered feasible; when the turning angle is too large, further verification is required. In this embodiment, the preset turning angle threshold θ_min = 30° (this value can be calibrated on-site according to the AGV model).

[0048] The second step is constraint judgment and elimination. For example, traverse all topological edges, and if its turning angle θ_ij < θ_min, mark the edge as impassable and remove it from the set of feasible edges.

[0049] S24: Based on the maximum travel speed, the estimated speed is obtained by evaluating the travel time of the remaining topological edges, and topological edges whose estimated speed exceeds the maximum travel speed are eliminated. The purpose of this step is to eliminate paths that the AGV cannot safely pass through within the specified time. The specific implementation steps are as follows: Step 1, Speed ​​Calculation: For the remaining topological edge e_ij, with length d_ij, based on the task time window issued by the production scheduling system, determine the maximum allowable time t_ij for the AGV to traverse the topological edge, and calculate the required travel speed V_ij = d_ij / t_ij. Here, V_ij is the minimum travel speed required by the AGV on the topological edge e_ij to meet the task time window, d_ij is the edge length (in meters), and t_ij is the maximum allowable travel time (in seconds).

[0050] The second step is constraint judgment and elimination: If V_ij > Vmax (the maximum speed of the AGV), then it is determined that the edge cannot be safely passed within the task time, and the edge is marked as exceeding the speed limit and eliminated.

[0051] S25: Redundancy removal is performed on the nodes and edges in the filtered topological network, deleting duplicate nodes and redundant edges to obtain the feasible topological network. The purpose of this step is to simplify the network size and improve the efficiency of subsequent searches. The specific implementation steps are as follows: The first step is to identify redundant nodes, for example, by using a jump point search algorithm to identify and delete intermediate nodes whose path direction has not changed.

[0052] The second step is to output the road network, such as a feasible topology road network after physical constraint filtering and redundancy reduction. The number of nodes and topology edges is reduced, but all physically feasible path options are retained.

[0053] By using constraint filtering based on AGV physical parameters, this invention ensures that each segment of the planned path meets the AGV's kinematic capabilities. This method solves the problem of planned paths being unexecutable due to infeasible turns or failure to meet speed requirements, thereby improving path executability from the source and reducing the risk of AGV jamming or safety accidents during execution.

[0054] In some embodiments, reference Figure 4 Step S30 includes: S31: Set the multi-dimensional weighting coefficients of the search algorithm, including a first coefficient for calculating turning costs, a second coefficient for calculating process costs, and a third coefficient for calculating safety costs. The specific implementation steps are as follows: The first step is to configure the coefficients. For example, for a warehousing scenario, the weight coefficients can be obtained through offline optimization using a genetic algorithm: for example, the turning cost weight α=0.3, the process cost weight β=0.6, and the safety cost weight γ=0.1.

[0055] The second step is coefficient application, for example, inputting the configured coefficients into the heuristic function calculation module of the search algorithm.

[0056] S32: Based on the aforementioned process chain, the global task is decomposed into multiple sub-path search tasks, with each sub-path starting and ending at a point that is a necessary step in an adjacent process. The specific implementation steps are as follows: The first step is task decomposition. For example, for a process chain C = [p1, p2, p3, p4], it is decomposed into three sub-tasks: searching for the optimal path from p1 to p2, from p2 to p3, and from p3 to p4.

[0057] The second step is to generate a task queue, for example, by placing subtasks into the search queue in order.

[0058] S33: For each sub-path search task, starting from the starting point, calculate the total cost of each candidate node based on a multi-constraint weighted heuristic function. The purpose of this step is to comprehensively evaluate the merits of each path option. The specific implementation steps are as follows: The first step is to calculate the heuristic function. For example, for node n, calculate its total cost f(n) = g(n) + h(n) + α·turn(n) + β·weight(n) + γ·safe(n). Here, g(n) is the actual distance from the starting point to n, h(n) is the Euclidean distance from n to the end point of the subtask, turn(n) = |θ_n| / 180°, θ_n is the turning angle of the current node n (in degrees), the turning angle is normalized to the interval [0,1], weight(n) is the process weight of the topological edge, safe(n) is the safety cost based on the obstacle distance, α is the turning cost weight, β is the process cost weight, and γ is the safety cost weight.

[0059] The second step is cost update, for example, storing the calculated f(n) value into the node's attributes.

[0060] S34: Guide the search based on the total cost until the end of the sub-path is reached, obtaining the optimal sub-path. The purpose of this step is to find the sub-path with the minimum cost. The specific implementation steps are as follows: The first step is node selection, for example, selecting the node with the smallest f(n) value from the open list as the current node for expansion.

[0061] The second step is path backtracking. For example, when expanding to the end point of a subtask, backtracking through the parent node pointer yields the optimal sub-path node sequence from the starting point to the end point.

[0062] S35: Concatenate all the optimal sub-paths according to the process chain sequence to generate the initial global optimal path. The purpose of this step is to synthesize a complete global path. The specific implementation steps are as follows: The first step is path concatenation, for example, connecting the node sequences of subpaths p1→p2, p2→p3, and p3→p4 in order, and removing duplicate connection points (such as p2 and p3).

[0063] The second step is to output the global path, for example, outputting the complete initial global optimal path node sequence, such as [p1, v1, p2, v2, p3, v3, p4].

[0064] By employing a heuristic function that integrates multiple constraints for process-oriented global search, this invention achieves comprehensive optimization of distance, turns, process priority, and safety in path planning. This method addresses the problem of traditional algorithms having a single optimization objective and failing to balance multiple requirements, thus generating a global path that achieves the optimal balance between length, smoothness, process compliance, and safety.

[0065] In some embodiments, reference Figure 5 Step S40 includes: S41: Acquire raw environmental perception data collected by the sensors mounted on the AGV. The purpose of this step is to obtain real-time environmental perception information. The specific implementation steps are as follows: The first step is data acquisition, for example, by using a lidar (such as SICK TIM571) mounted on the AGV to scan the surrounding environment at a frequency of 10Hz to obtain distance point cloud data; at the same time, environmental images are acquired through an industrial camera.

[0066] The second step is data preprocessing, such as filtering and denoising point cloud data, and correcting image distortion.

[0067] S42: Perform data fusion processing on the raw environmental perception data to identify dynamic obstacles in the environment. The purpose of this step is to accurately identify and locate moving obstacles in the environment. The specific implementation steps are as follows: The first step is data fusion, such as spatiotemporal registration and fusion of LiDAR point clouds and visual images to construct a unified representation of the environment.

[0068] The second step is obstacle recognition, which involves using deep learning models to identify dynamic obstacles such as pedestrians, other AGVs, and temporary stacked objects from the fused data, and outputting their category labels.

[0069] S43: Extract the position coordinates, speed, and direction of motion of the dynamic obstacle. The purpose of this step is to quantify the obstacle's motion state for collision prediction. The specific implementation steps are as follows: The first step is state estimation, for example, using the Kalman filter algorithm to track the identified dynamic obstacles and estimate their position, velocity vector, and direction of motion in the world coordinate system.

[0070] The second step is information encapsulation, such as encapsulating obstacle ID, location, speed, direction, and other information into structured obstacle information objects.

[0071] S44: Identify whether the dynamic obstacle has entered the AGV's preset sensing range and is affecting the current travel path. The purpose of this step is to determine whether the obstacle affects the path. The specific implementation steps are as follows: The first step is range determination, such as calculating the Euclidean distance between the obstacle and the AGV, and determining whether it is less than the preset sensing range (e.g., 5 meters).

[0072] The second step is to determine the impact of the path, such as predicting the trajectory of an obstacle in the next few seconds and determining whether it overlaps with the spatiotemporal envelope of the AGV's current planned path.

[0073] S45: When the identification result is yes, the obstacle information is acquired and a warning signal is generated; wherein, the obstacle information includes the position coordinates, speed, and direction of movement of the dynamic obstacle. The purpose of this step is to trigger the obstacle avoidance response process. The specific implementation steps are as follows: The first step is information transfer, such as passing the complete obstacle information object to the dynamic replanning module.

[0074] The second step is signal generation, such as simultaneously generating a high-priority warning signal to notify the AGV control system to prepare to receive new path instructions.

[0075] By fusing LiDAR and visual sensors, this invention achieves accurate identification and state estimation of dynamic obstacles. This method solves the problems of insufficient detection accuracy and susceptibility to missed or false detections caused by single sensors in complex lighting or occlusion environments, thus providing reliable environmental perception input for subsequent local replanning and ensuring the accuracy of obstacle avoidance decisions.

[0076] In some embodiments, reference Figure 6 Step S50 includes: S51: Based on the obstacle information, locate the affected topological edges in the feasible topological road network. The purpose of this step is to determine the road network area that needs adjustment. The specific implementation steps are as follows: The first step is spatial mapping, such as mapping the location coordinates of obstacles to the topological road network and finding the nearest topological edge.

[0077] The second step is impact assessment, such as determining whether the movement trajectory of the obstacle covers or is adjacent to the topological edge. If so, the topological edge is marked as an affected topological edge.

[0078] S52: Mark the affected topological edges as temporarily impassable and adjust their technological and safety costs. The purpose of this step is to increase the passage cost of the affected area in the road network, guiding the search to avoid that area. The specific implementation steps are as follows: The first step is to set a status flag, for example, by temporarily setting the "walkable" attribute of the affected topological edge to "false".

[0079] The second step is to adjust the cost. For example, the process weight (n) of the edge is temporarily set to the maximum value of 1.0, and the safety cost (safe(n)) is recalculated and increased based on the distance to the obstacle.

[0080] S53: Using a preset number of nodes before and after the affected road segment as the start and end points of the local search, determine the local replanning area. The purpose of this step is to limit the scope of replanning and avoid global calculations. The specific implementation steps are as follows: The first step is to define the region. For example, on the initial global path, take 5 nodes forward and 5 nodes backward, centered on the affected edge, as the starting and ending points of the local replanning.

[0081] The second step is subgraph extraction, for example, extracting a local subgraph containing these nodes and related edges from the feasible topological network, as the search space for replanning.

[0082] S54: Within the local replanning region, a local replanning heuristic function is used for path search. The purpose of this step is to find the optimal detour path within a local area. The specific implementation steps are as follows: The first step is to define the heuristic function. For example, the local reprogramming heuristic function can be defined as f_local(n) = g_local(n) + h_local(n) + α·turn(n) + β·weight(n) + γ·safe(n) + δ·conflict(n). Here, conflict(n) is the probability of conflict between node n and dynamic obstacles, with a value range of [0,1], which is predicted from the position, speed and direction of movement of the obstacle; δ is the conflict weight coefficient, with a value range of [0.2, 0.4].

[0083] The second step is local search. For example, in a local subgraph, the search algorithm is run to search using the starting point and ending point determined in step S53 to obtain the local optimal path.

[0084] S55: Replace the affected road segments in the initial global optimal path with the locally optimal path obtained from the search, and then concatenate them to generate the final executable path. The purpose of this step is to update the global path. The specific implementation steps are as follows: The first step is path replacement, for example, replacing the original road segment from the local starting point to the end point in the initial global path with the locally optimal path obtained in step S54.

[0085] The second step is to output the path, for example, output the updated complete path as the final executable path, and send it to the AGV execution module.

[0086] This invention achieves rapid response to dynamic obstacles through a local replanning mechanism. This method solves the problems of large computational load and long processing time associated with global replanning, thereby minimizing the impact on AGV operation continuity while ensuring obstacle avoidance safety, and improving the system's real-time performance and robustness in dynamic environments.

[0087] In some embodiments, reference Figure 7 Step S60 includes: S61: Extract a series of key path points as control points from the final executable path. The purpose of this step is to select key nodes for curve fitting. The specific implementation steps are as follows: The first step is point sequence extraction, which involves extracting the coordinates of all nodes sequentially from the node sequence of the final executable path.

[0088] The second step is to determine the control points, for example, by directly using the extracted path points as the control points P_i of the cubic B-spline curve.

[0089] S62: Using a cubic B-spline curve algorithm, path smoothing is performed based on the control points to obtain a smoothed continuous curve. The purpose of this step is to generate a smooth trajectory with continuous curvature. The specific implementation steps are as follows: The first step is to calculate the basis functions. For example, the smoothing formula for a cubic B-spline curve is: ; Where B(t) is a cubic B-spline basis function, specifically: ; Where P is the path control point (the key node of the planned path), and t∈[0,1].

[0090] The second step is curve generation. Based on the smoothing formula of cubic B-spline curves, a smooth curve segment is generated for every four adjacent control points to obtain the entire smooth path curve.

[0091] S63: Discretize the smoothed continuous curve to obtain a series of dense path point sequences. The purpose of this step is to transform the continuous curve into discrete control target points. The specific implementation steps are as follows: The first step is parameter sampling, for example, sampling at fixed intervals Δt within the parameter t ∈ [0,1].

[0092] The second step is to generate a point sequence. For example, each sampling parameter t_i is substituted into the smoothing formula of the cubic B-spline curve to obtain a series of dense path point coordinates.

[0093] S64: Based on the path point sequence and the AGV's kinematic model, calculate the linear velocity and angular velocity commands corresponding to each sampling moment. The purpose of this step is to convert the geometric path into motion control commands. The specific implementation steps are as follows: The first step is kinematic modeling, such as establishing a differential drive kinematic model for an AGV.

[0094] The second step involves instruction calculation, such as calculating the desired linear velocity V based on the distance and time interval between adjacent path points, and calculating the desired angular velocity ω based on the curvature at the path points.

[0095] S65: The linear velocity and angular velocity commands are sent to the AGV controller to drive the AGV to execute. The purpose of this step is to control the AGV to track a smooth path. The specific implementation steps are as follows: The first step is instruction encapsulation, for example, encapsulating the calculated linear velocity V and angular velocity ω into control instruction frames.

[0096] The second step is command sending and execution. For example, command frames are sent to the AGV's underlying controller via CAN bus or Ethernet. The controller drives the motor to execute the command, allowing the AGV to travel smoothly along a smooth path.

[0097] By combining cubic B-spline curve smoothing with kinematic command conversion, this invention achieves the transformation from discrete topological paths to continuous smooth motion trajectories. This method solves the problems of motion jerking and large turning impacts caused by directly tracking polyline paths in AGVs, thereby improving the stability and accuracy of AGV path tracking and reducing mechanical wear.

[0098] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0099] Reference Figure 8 This invention provides a multi-constraint fusion AGV path planning device 100, comprising: The topology network construction module 101 is used to construct a topology network containing production process constraints, resulting in a topology network containing mandatory nodes. The constraint filtering module 102 is used to perform constraint filtering on the topology network based on the physical parameters of the AGV to obtain a feasible topology network. The path search module 103 is used to perform a process-oriented global path search in the feasible topology network using a search algorithm to generate an initial global optimal path; The dynamic obstacle detection module 104 is used to collect environmental data in real time, perform dynamic obstacle detection based on the environmental data, and obtain obstacle information. The dynamic replanning module 105 is used to perform local replanning on the initial global optimal path based on the obstacle information to generate a final executable path. The path execution module 106 is used to smooth the final executable path and convert it into control instructions, and drive the AGV to execute through the control instructions.

[0100] Specific limitations regarding the multi-constraint fusion AGV path planning device 100 can be found in the limitations of the multi-constraint fusion AGV path planning method described above, and will not be repeated here. Each module in the aforementioned multi-constraint fusion AGV path planning device 100 can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device in hardware form, or stored in the memory of a computer device in software form, so that the processor can call and execute the operations corresponding to each module.

[0101] This invention also provides a computer device, which includes a processor, a memory, a network interface, and a database connected via a system bus. The processor provides computational and control capabilities. The memory includes non-volatile and / or volatile storage media and internal memory. The non-volatile storage media stores an operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The network interface communicates with external clients via a network connection. When the computer program is executed by the processor, it implements the functions or steps of the multi-constraint fusion AGV path planning method provided in the above embodiments.

[0102] This invention also provides a computer-readable storage medium storing a computer program, which, when executed, can implement the functions or steps of the multi-constraint fusion AGV path planning method provided in the above embodiments.

[0103] It should be noted that the functions or steps that can be implemented by the computer-readable storage medium or computer device described above can be referred to the relevant descriptions on the server side and client side in the foregoing method embodiments. To avoid repetition, they will not be described one by one here.

[0104] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0105] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0106] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A multi-constraint fusion AGV path planning method, characterized in that, include: Construct a topology network that includes production process constraints to obtain a topology network that includes mandatory nodes; The topological network is constrained and filtered based on the physical parameters of the AGV to obtain a feasible topological network. A search algorithm is used to perform a process-oriented global path search in the feasible topology network to generate an initial global optimal path; Real-time environmental data is collected, and dynamic obstacle detection is performed based on the environmental data to obtain obstacle information; Based on the obstacle information, the initial global optimal path is locally replanned to generate the final executable path; The final executable path is smoothed and converted into control instructions, which are then used to drive the AGV to execute.

2. The AGV path planning method with multi-constraint fusion according to claim 1, characterized in that, The construction of a topology network including production process constraints, resulting in a topology network including mandatory nodes, includes: Acquire workshop environment data, generate a raster map based on the workshop environment data, and extract key nodes from the raster map to construct an initial topology node set; Establish topological edges connecting adjacent nodes in the initial topological node set, and set distance attributes, turning angle attributes, and process weight attributes for each topological edge; Analyze production process data, extract the set of necessary process points, and generate process chains according to the process sequence; Map each point in the set of necessary process points to the grid map and identify them as mandatory nodes. Merge the mandatory nodes, the initial set of topological nodes, and the topological edges to obtain the topological road network.

3. The AGV path planning method with multi-constraint fusion according to claim 1, characterized in that, The constraint filtering of the topology network based on the physical parameters of AGVs to obtain a feasible topology network includes: Obtain the physical parameters of the AGV, including the minimum turning radius and the maximum travel speed; Traverse all topological edges in the given topological network and calculate the turning angle corresponding to each topological edge; The turning angle of each topological edge is compared with the minimum feasible turning angle calculated based on the minimum turning radius, and topological edges with turning angles smaller than the minimum feasible turning angle are removed. Based on the maximum driving speed, the travel time of the remaining topological edges is evaluated to obtain the estimated speed, and topological edges whose estimated speed exceeds the maximum driving speed are removed. Redundancy removal is performed on the nodes and edges in the filtered topological network, deleting duplicate nodes and redundant edges to obtain the feasible topological network.

4. The AGV path planning method with multi-constraint fusion according to claim 2, characterized in that, The step of employing a search algorithm to perform a process-oriented global path search within the feasible topology network to generate an initial globally optimal path includes: The search algorithm is configured with multi-dimensional weighting coefficients, including a first coefficient for calculating turning costs, a second coefficient for calculating process costs, and a third coefficient for calculating safety costs. Based on the aforementioned process chain, the global task is decomposed into multiple sub-path search tasks with the necessary points of adjacent processes as the starting and ending points. For each sub-path search task, starting from the starting point, the total cost of each candidate node is calculated based on a heuristic function with multiple constraints and weights. The search is guided by the total cost until the end of the sub-path is reached, thus obtaining the optimal sub-path. By assembling all the optimal sub-paths in the process chain sequence, the initial global optimal path is generated.

5. The AGV path planning method with multi-constraint fusion according to claim 1, characterized in that, The real-time acquisition of environmental data for dynamic obstacle detection, to obtain obstacle information, includes: Acquire raw environmental perception data collected by the sensors mounted on the AGV; The raw environmental perception data is fused to identify dynamic obstacles in the environment; Extract the position coordinates, speed, and direction of motion of the dynamic obstacle; Identify whether the dynamic obstacle has entered the AGV's preset sensing range and affected the current travel path; When the identification result is yes, the obstacle information is obtained and a warning signal is generated; wherein, the obstacle information includes the position coordinates, movement speed and movement direction of the dynamic obstacle.

6. The AGV path planning method with multi-constraint fusion according to claim 1, characterized in that, The step of performing local replanning on the initial globally optimal path based on the obstacle information to generate a final executable path includes: Based on the obstacle information, locate the affected topological edge in the feasible topological road network; Mark the affected topological edges as temporarily impassable and adjust their technological and security costs accordingly. The local replanning area is determined by using a preset number of nodes before and after the affected road segment as the starting and ending points of the local search; Within the local replanning region, a local replanning heuristic function is used for path search; The affected road segments in the initial global optimal path are replaced by the locally optimal path obtained from the search, and then the resulting paths are combined to generate the final executable path.

7. The AGV path planning method with multi-constraint fusion according to claim 1, characterized in that, The process of smoothing the final executable path and converting it into control instructions, and then driving the AGV to execute the path using these control instructions, includes: Extract a series of critical path points as control points from the final executable path; A cubic B-spline curve algorithm is used to smooth the path based on the control points, resulting in a smoothed continuous curve. Discretize the smoothed continuous curve to obtain a series of dense path point sequences; Based on the path point sequence and the kinematic model of the AGV, calculate the linear velocity command and angular velocity command corresponding to each sampling moment; The linear velocity command and angular velocity command are sent to the AGV controller to drive the AGV to execute.

8. A multi-constraint fusion AGV path planning device, characterized in that, include: The topology network construction module is used to construct a topology network that includes production process constraints, resulting in a topology network containing mandatory nodes. The constraint filtering module is used to perform constraint filtering on the topological road network based on the physical parameters of the AGV to obtain a feasible topological road network. The path search module is used to perform a process-oriented global path search in the feasible topology network using a search algorithm to generate an initial global optimal path. The dynamic obstacle detection module is used to collect environmental data in real time, perform dynamic obstacle detection based on the environmental data, and obtain obstacle information. The dynamic replanning module is used to perform local replanning on the initial global optimal path based on the obstacle information to generate the final executable path. The path execution module is used to smooth the final executable path and convert it into control instructions, and drive the AGV to execute through the control instructions.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the AGV path planning method with multi-constraint fusion as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the multi-constraint fusion AGV path planning method as described in any one of claims 1 to 7.

Images